Transcript Probabilistic Information Retrieval
Ahmet Selman Bozkır
Introduction to conditional, total probability & Bayesian theorem Historical background of probabilistic information retrieval Why probabilities in IR?
Document ranking problem Binary Independence Model
Given some event B with nonzero probability P(B) > 0 We can define conditional prob. as an event A, given B, by
P
(
A B
)
P
(
A
B
)
P
(
B
) The Probabilty P(A|B) simply reflects the fact that the probability of an event A may depend on a second event B. So if A and B are mutually exclusive, A B =
Resistance ( ) 22 47 100 Total:
Tolerance 5% 10 28 24 62 10% 14 26 8 38 Total
The joint probabilities are: P(A B) = P(47 5%) = 28/100 P(A P(B C) = P(47 C) = P(5% 100 ) = 0 100 ) = 24/100
24 44 32 100
Let’s define three events: 1. A as “draw 47 resistor 2. B as “draw” a resistor with 5% 3. C as “draw” a “100 P(A) = P(47 P(B) = P(5%) = 62/100 P(C) = P(100 I f we use them the cond. prob. :
P
(
A B
)
P
(
A
B
)
P
(
B
) 28 62 ) = 44/100 ) = 32 /100
P
(
A C
)
P
(
B C
) resistor
P
(
A
C
)
P P
(
B P
( (
C
C
)
C
) ) 0 24 32
The probability of P(A) of any event A defined on a sample space S can be expressed in terms of cond. probabilities. Suppose we are given N mutually exclusive events B
n
ilustrated in figure ,n = 1,2…. N whose union equals S as A B n B1 B2 A A S A
n N
1
B n
n N
1 (
A
B n
) B3 Bn
The definition of conditional probability applies to any two events. In particular ,let B n be one of the events defined above in the subsection on total probability.
P
(
B n A
)
P(B n
A) P(A)
İf P(A)≠O,or, alternatively,
P
(
A B n
)
P
(
A
B n
)
P
(
B n
)
if P(B n )≠0, one form of Bayes’ theorem is obtained by equating these two expressions:
P
(
B n A
)
P
(
A B n
)
P
(
B n
)
P
(
A
) Another form derives from a substitution of P(A) as given:
P
(
B n A
)
P
(
A P
(
B
1 )
P
(
B
1 )
A
B n
...
)
P
(
P B n
(
A
)
B N
)
P
(
B N
)
The first attempts to develop a probabilistic theory of retrieval were made over 30 years ago [Maron and Kuhns 1960; Miller 1971], and since then there has been a steady development of the approach. There are already several operational IR systems based upon probabilistic or semiprobabilistic models.
One major obstacle in probabilistic or semiprobabilistic IR models is finding methods for estimating the probabilities used to evaluate the probability of relevance that are both theoretically sound and computationally efficient.
The first models to be based upon such assumptions were the “binary independence indexing model” and the “binary independence retrieval model One area of recent research investigates the use of an explicit network representation of dependencies. The networks are processed by means of Bayesian inference or belief theory, using evidential reasoning techniques such as those described by Pearl 1988. This approach is an extension of the earliest probabilistic models, taking into account the conditional dependencies present in a real environment.
User Information Need Query Representation
Understanding of user need is uncertain
How to match?
Document Representation
Uncertain guess of whether document has relevant content
Document s
In traditional IR systems, matching between each document and query is attempted in a semantically imprecise space of index terms.
Probabilities provide a principled foundation for uncertain reasoning.
Can we use probabilities to quantify our uncertainties?
Classical probabilistic retrieval model Probability ranking principle, etc.
(Naïve) Bayesian Text Categorization Bayesian networks for text retrieval
Probabilistic methods are one of the oldest but also one of the currently hottest topics in IR.
Traditionally: neat ideas, but they’ve never won on performance. It may be different now.
In probabilistic information retrieval, the goal is the estimation of the probability of relevance P(R l q
k , d m
) that a document d m will be judged relevant by a user with request q k . In order to estimate this probability, a large number of probabilistic models have been developed. Typically, such a model is based on representations of queries and documents (e.g., as sets of terms); in addition to this, probabilistic assumptions about the distribution of elements of these representations within relevant and nonrelevant documents are required.
By collecting relevance feedback data from a few documents, the model then can be applied in order to estimate the probability of relevance for the remaining documents in the collection.
We have a collection of documents User issues a query A list of documents needs to be returned
Ranking method is core of an IR system:
In what order do we present documents to the user?
We want the “best” document to be first, second best second, etc….
Idea: Rank by probability of relevance of the document w.r.t. information need
P(relevant|document i , query)
For events a and b:
p
Bayes’ Rule
(
a
,
b
)
p
(
a
b
)
p
(
a
|
b
)
p
(
b
)
p
(
b
|
a
)
p
(
a
)
p
(
a p
(
a
|
b
)
p
(
b
) |
b
) Posterior
p
(
b
|
p
(
b
|
a
)
p
(
a
)
a
)
p
(
a
)
p
(
b
)
p
(
b
x
a
,
a
|
a
)
p
(
b
|
p
(
a
)
x
)
p
(
x
) Prior
Odds:
O
(
a
)
p
(
a p
(
a
) ) 1
p
(
a
)
p
(
a
)
Let
x
be a document in the collection. Let
R
represent
relevance
of a document w.r.t. given (fixed) query and let
NR
represent
non-relevance.
R={0,1} vs. NR/R Need to find p(
R|x)
- probability that a document
x
is
relevant.
p
(
R
|
x
)
p
(
x
|
R
)
p
(
R
)
p
(
x
)
p
(
NR
|
x
)
p
(
x
|
NR
)
p
(
NR
)
p
(
x
) p(
R)
,p(
NR
) - prior probability of retrieving a (non) relevant document
p
(
R
|
x
)
p
(
NR
|
x
) 1 p(
x|R
), p(
x|NR
)
-
probability that if a relevant (non-relevant) document is retrieved, it is
x .
Bayes’ Optimal Decision Rule
x
is relevant iff p(R|x) > p(NR|x)
PRP in action: Rank all documents by p(R|x)
More complex case: retrieval costs.
Let d be a document C - cost of retrieval of relevant document C’ - cost of retrieval of non-relevant document
C
Probability Ranking Principle: if
p
(
R
|
d
)
C
( 1
p
(
R
|
d
))
C
p
(
R
|
d
)
C
( 1
p
(
R
|
d
))
for all d’ not yet retrieved, then d is the next
document to be retrieved
We won’t further consider loss/utility from now on
How do we compute all those probabilities?
Do not know exact probabilities, have to use estimates Binary Independence Retrieval (BIR) – which we discuss later today – is the simplest model
Questionable assumptions
“Relevance” of each document is independent of relevance of other documents.
▪ Really, it’s bad to keep on returning duplicates Boolean model of relevance
Estimate how terms contribute to relevance
How tf, df, and length influence your judgments about do things like document relevance? ▪ One answer is the Okapi formulae (S. Robertson)
Combine to find document relevance probability
Order documents by decreasing probability
Basic concept:
"For a given query, if we know some documents that are relevant, terms that occur in those documents should be given greater weighting in searching for other relevant documents.
By making assumptions about the distribution of terms and applying Bayes Theorem, it is possible to derive weights theoretically."
Van Rijsbergen
Traditionally used in conjunction with PRP
“Binary” = Boolean
: documents are represented as binary incidence vectors of terms (cf. lecture 1):
x
(
x
1 , ,
x n
)
x i
1 iff term i is present in document x.
“Independence”:
terms occur in documents independently Different documents can be modeled as same vector Bernoulli Naive Bayes model (cf. text categorization!)
Queries: binary term incidence vectors Given query
q
, for each document
d
need to compute
p
(
R
|
q,d
)
.
replace with computing
p
(
R
|
q,x
)
where
x
is binary term incidence vector representing
d
Interested only in ranking Will use odds and Bayes’ Rule:
O
(
R
|
q
,
x
)
p
(
R p
(
NR
| |
q
,
q
,
x
x
) )
p
(
R p
(
NR
| |
q
)
p
(
x p
( |
q
)
p
(
x p
( |
x q x
) |
q
) |
R
,
q
)
NR
,
q
)
O
(
R
|
q
,
x
)
p
(
R p
(
NR
| |
q
,
q
,
x
x
) )
p
(
R
|
p
(
NR
|
q
)
q
)
p p
( (
x x
| |
R
,
q
)
NR
,
q
) Constant for a given query Needs estimation • Using
Independence
Assumption:
p p
( (
x
x
| |
R
,
q
)
NR
,
q
)
i n
1
p p
( (
x x i
|
i
|
R
,
q
)
NR
,
q
) •So :
O
(
R
|
q
,
d
)
O
(
R
|
q
)
i n
1
p
(
x i p
(
x i
|
R
,
q
) |
NR
,
q
)
O
(
R
|
q
,
d
)
O
(
R
|
q
)
i n
1
p
(
x i
|
R
,
q
)
p
(
x i
|
NR
,
q
) • Since
x i
is either 0 or 1:
O
(
R
|
q
,
d
) • Let
p i
O
(
R
|
p
(
x i q
)
x i
1 1 |
R
,
q
);
p p
( (
x i x i
1 1 | |
R
,
NR
,
q
)
q
)
x i
0
r i
p
(
x i p p
( (
x i x i
0 0 | |
R
,
q
)
NR
,
q
) 1 |
NR
,
q
); • Assume, for all terms not occurring in the query (
q i =0
) Then...
p i
r i
This can be changed (e.g., in relevance feedback)
O
(
R
|
q
,
x
)
O
(
R
|
q
)
x i
q i
1 All matching terms
p i r i
O
(
R
|
q
)
x i
q i
1
r i p i
( 1 ( 1
r p i i
) )
x q i i
0 1 1 1
p i r i
Non-matching
q i
1 1 1 query terms
p i r i
All matching terms All query terms
O
(
R
|
q
,
x
)
O
(
R
|
q
)
x i
q i
1
r i p i
( 1 ( 1
r p i i
) )
q i
1 1 1
p i r i
Constant for each query Only quantity to be estimated for rankings • Retrieval Status Value:
RSV
log
x i
q i
1
r i p i
( 1 ( 1
r p i i
) )
x i
q i
1 log
r i p i
( 1 ( 1
r p i i
) )
• Estimating RSV coefficients.
• For each term
i
look at this table of document counts: Documens Relevant Non-Relevant Total
X i =1 X i =0
s S-s n-s N-n-S+s n N-n
Total
S N-S N c i
• Estimates:
p i
s S
K
(
N
,
n
,
S
,
s
) log
r i
(
n
( (
n
s
)
N
S
)
s
)
s
( (
S N
n s
)
S
s
) For now, assume no zero terms.
If non-relevant documents are approximated by the whole collection, then r
i
(prob. of occurrence in non-relevant documents for query) is n/N and
log (1 – r
i
)/
r i
= log (N – n )/
n
≈ log N/
n
= IDF!
p i
(probability of occurrence in relevant documents) can be estimated in various ways: from relevant documents if know some ▪ Relevance weighting can be used in feedback loop constant (Croft and Harper combination match) – then just get idf weighting of terms proportional to prob. of occurrence in collection ▪ more accurately, to log of this (Greiff, SIGIR 1998)
1.
2.
3.
4.
Assume that p
p i i
constant over all x
i
= 0.5 (even odds) for any given doc in query Determine guess of relevant document set: V is fixed size set of highest ranked documents on this model (note: now a bit like tf.idf!) We need to improve our guesses for p
i
and r
i
, so ▪ Use distribution of x
i
in docs in V. Let V i documents containing x
i p i
= |V i | / |V| be set of Assume if not retrieved then not relevant ▪
r i
= (n i – |V i |) / (N – |V|) Go to 2. until converges then return ranking
1.
2.
3.
4.
Guess a preliminary probabilistic description of R and use it to retrieve a first set of documents V, as above.
Interact with the user to refine the description: learn some definite members of R and NR Reestimate p
i
and r
i
on the basis of these Or can combine new information with original guess (use Bayesian prior):
p i
( 2 ) |
V
|
i V
| | Repeat, thus generating a succession of
p
i
( 1 ) κ is prior weight approximations to R.
Getting reasonable approximations of probabilities is possible.
Requires restrictive assumptions:
term independence
terms not in query don’t affect the outcome
boolean representation of documents/queries/relevance
document relevance values are independent
Some of these assumptions can be removed Problem: either require partial relevance information or only can derive somewhat inferior term weights
In general, index terms aren’t independent Dependencies can be complex van Rijsbergen (1979) proposed model of simple tree dependencies Exactly Friedman and Goldszmidt’s Tree Augmented Naive Bayes (AAAI 13, 1996) Each term dependent on one other In 1970s, estimation problems held back success of this model
What is a Bayesian network?
A directed acyclic graph Nodes ▪ Events or Variables ▪ Assume values. ▪ For our purposes, all Boolean Links ▪ model direct dependencies between nodes
p(a) a c b p(c|ab)
for all values for
a,b,c
• Bayesian networks model causal relations between events
p(b)
Conditional •Inference in Bayesian Nets: •Given probability distributions for roots and conditional dependence probabilities can compute apriori
probability
of any instance • Fixing assumptions (e.g.,
b
was observed) will cause recomputation of probabilities For more information see: R.G. Cowell, A.P. Dawid, S.L. Lauritzen, and D.J. Spiegelhalter.
1999.
Probabilistic Networks and Expert Systems
. Springer Verlag.
J. Pearl. 1988.
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference.
Morgan-Kaufman.
f f
0 .
3 0 .
7 Finals (f)
n
n f
0 .
9 0 .
1
f
0 .
3 0 .
7 No Sleep (n)
t
t g
0 .
99 0 .
01
g
0 .
1 0 .
9 Project Due (d) Triple Latte (t)
d
d
0 .
4 0 .
6 Gloom (g)
g
g fd
0 .
99 0 .
01
fd
0 .
9 0 .
1
f
d
0 .
8 0 .
2
f
d
0 .
3 0 .
7
Finals (f) Project Due (d) No Sleep (n) Gloom (g) Triple Latte (t) • Independence assumption: P(t|g, f)=P(t|g) • Joint probability P(f d n g t) =P(f) P(d) P(n|f) P(g|f d) P(t|g)
Goal
Given a user’s information need (evidence), find probability a doc satisfies need
Retrieval model
Model docs in a document network Model information need in a query network
Document Network d1 d2 d
i -
documents t1 r1 t2 r2 t
i
- document representations r - “concepts” for each document collection r3 r
k
d
n
t
n
c1 c2 q
i -
q1 Query Network c
i
- query concepts Small, compute once for query q2 c
m
I I - goal node
Construct Document Network (once !) For each query
Construct best Query Network Attach it to Document Network Find subset of
d
i
’s
which maximizes the probability value of node
I
(best subset).
Retrieve these
d
i
’s
as the answer to query.
r 1 d 1 Documents r 2 d 2 r 3 Terms/Concepts
Document Network
c 1 q 1 c 2 c 3 Concepts q 2 Query operators (
AND/OR/NOT
) i Information need
Query Network
Prior doc probability P(d) = 1/n P(r|d) within-document term frequency
tf
idf - based P(c|r) 1-to-1 thesaurus P(q|c): canonical forms of query operators Always use things like AND and NOT – never store a full CPT* *conditional probability table
Hamlet reason Macbeth trouble double reason trouble
OR
User query
NOT
two
Document Network Query Network
Prior probs don’t have to be 1/n.
“User information need” doesn’t have to be a query - can be words typed, in docs read, any combination … Phrases, inter-document links Link matrices can be modified over time.
User feedback.
The promise of “personalization”
Document network built at indexing time Query network built/scored at query time Representation:
Link matrices from docs to any single term are like the postings entry for that term Canonical link matrices are efficient to store and compute
Attach evidence only at roots of network
Can do single pass from roots to leaves